instructor

You can extract structured data from LLM responses by defining a Pydantic model as your response schema, which validates the extracted output and automatically retries on validation failures to ensure correctness.

Yes, structured output extraction supports multi-provider compatibility across OpenAI, Anthropic, and local models, allowing you to apply the same Pydantic validation schemas regardless of the LLM provider.

When structured output fails Pydantic validation, the system applies automatic retry logic, prompting the LLM again to correct the formatting errors and ensure the response matches the defined schema.

Yes, it supports streaming partial results for real-time updates, allowing you to process multi-entity information and structured data progressively as the LLM generates the response tokens.

The best way to extract user profiles and classifications from unstructured text is using Pydantic schemas to map the LLM output into structured data, ensuring strict validation and automatic retries for malformed responses.